Refractory materials by definition are those used at the high temperatures required to activate processes essential to the ceramics, chemical, metallurgical and power generation industries.

Thermally insulating systems are necessary to confine these high temperature environments, in order to achieve safe and tolerable working environments, protect proximate structures and equipment, and maximise energy efficiency.

Thus, high temperature insulating refractories are a very important family of materials.

These materials are invariably porous to a very high degree, which confers low thermal conductivity because heat transfer through the pores is usually less than through the solid.

Such materials also have low thermal mass (the heat per unit volume for a unit change in temperature) because of their low solid content. In dynamic applications, such as cyclic heating processes and for kiln furniture moving through a hot zone, thermal mass can be as important as thermal conductivity. However, insulating refractories are a compromise as a consequence of their porosity, which impairs structural integrity and corrosion/erosion resistance.

Prior to the 1950's traditional insulating firebricks were the main insulating refractories.

Over the last 50 years many new high temperature insulation systems and materials have been developed and have provided significant benefits to the process industries. These developments led to the introduction of the high temperature fibrous insulation materials from the early 1950's, although they were not widely employed until the energy crises of the 1960's.

The economic benefits of low thermal mass systems, through energy saving in intermittent/cyclic temperature processing and lightweight constructions is now well documented (see Figures 1 and 2). These factors have enabled industry in N America and Europe to make savings currently estimated at US$2, 50OM/annum, and to reduce emissions of"greenhouse"gases by-15M tonnes/annum.

The choice of insulating material system facing designers and users of high temperature processes is considerable, with the final selection between competing products depending on the relative weighting given to technical performance, life-cycle cost, and acceptable health & safety and environmental risks. The inventors, by an analysis of these factors, have made the invention claimed herein.

Heat transfer through highly porous structures is complex and predictive models have to assume a uniformity of the microstructure that is usually absent in practice. This said, such models do provide a useful basis for understanding the contribution of the major parameters determining thermal properties.

The major contributing mechanisms of heat transfer in high temperature insulating refractories are : Conduction through the solid phase.

Conduction through the gas phase.

Radiation from the solid surfaces, usually through the gas phase but also through the solid phase if it is transparent at infra-red wavelengths.

Convection is another heat transfer mechanism in the gas phase but its contribution is negligible compared with radiative transfer in structures with pore sizes less than a few millimetres.

Most high temperature insulating refractories are based on oxide ceramics and glasses. With these materials as solids, the rate of heat transfer reduces with increasing temperature. The reduction is greater for materials with simple crystallographic structures, which commonly have higher thermal conductivity at room temperature, and less for materials with more complex arrangements of atoms or with disordered or disrupted microstructures, which have lower thermal conductivity.

For example, pure magnesia has a Kc (thermal conductivity in W/m. K) value of #50 at 20°C and #9 at 1000°C. By comparison, quartz and mullite have a KC values of #8, zircon of-5 and vitreous silica of-1. 5 at 20°C. The effective conductivity of a partially dense solid, at relatively low levels of porosity, may be approximated by Kcen-Vf. Kc, (1) where, Vf is the volume fraction of the solid.

The conductivity of gases may be derived from kinetic theory, and is a function of their specific heat, the mass of the gas molecules, their mean free path, and the number of molecules available for collision (e. g. as a function of pressure). Unlike most solids, conduction in gases typically increases by an order of magnitude between 0 and 1500°C.

For air, Kc is, approximately, Kc = 0. 0004. T/', (2) where T is absolute temperature, with a value of-0. 03W/m. K at 20°C and #0. 1 at 1500°C.

The mean free path, 1, is /=3. 5. 10-'. T (um), (3) i. e. #0.2 µm at 20°C and 0.6 µm at 1500°C. This parameter is important because if the gas molecules'movement is physically constrained, for example by containment in a void of dimension < 1, KC of the gas is reduced by a factor p/(l + p), (4) where p is the pore size.

In high temperature thermal environments radiation is the transfer of heat in the infra-red range. Stefan's law applies to a true black body, which is a perfect emitter and absorber, such that the heat radiated from the surface E = o. T o. r. AT = KR. A. AT/L (W/m2), (5) when the temperature difference between emitter and absorber, AT, is small, and where a is the Stefan-Boltzmann constant, 5. 7xl0-8W./mz. K4, A is the surface area, KR is the radiative thermal conductivity, and L is distance between emitting and absorbing surfaces. The T3 relationship means that at high temperatures radiation becomes the dominant heat transfer mechanism.

Refractory oxide surfaces do not approximate to black bodies because their absorptivity, a, and reflectivity, r, are <1, and determine emissivity, s, £=a= 1-r. (6) Reflectivity is related to the refractive index of the material, n, which for most oxides is between 1. 5 and 3, and to the wavelength of the radiation. It decreases, so emissivity increases, as refractive index and wavelength decrease, and, therefore, with increasing temperature according to Wien's formula for the peak radiation wavelength, ?,, k = 0. 0029/T (m). (7) Since thermal insulating refractories are composite materials of solid and gas, predictive thermal conductivity models need to account for the properties and content of individual components, their geometry and the contribution of different thermal transfer mechanisms.

Specific models have been built for : "Porous"structures, defined as gas pores dispersed in a solid, and "Particulate"structures, defined as solid fibres, platelets (flakes) and spheres dispersed in a gas.

For the former, it might appear from relative Kc values that significantly reduced thermal conductivity would be achieved by simply increasing the void size and fraction. But, when high temperature radiative transfer is taken into account this is not the case. If equations (5) and (6) are applied to the emission and absorption of heat from the walls of a small cavity, we obtain an effective radiative thermal conductivity KReff, KReff=. T\p, (8) where kl combines the Stefan-Boltzmann constant and emissivity, with a value of #9x10- 8W/m2. K4, and p is the cavity (pore) diameter in m. Equation (8) predicts values for KReff at 1500°C of-0. 0025W/m. K for pores of 5Rm diameter,-0. 025W/m. K for pores of 50pm diameter and-2. 5W/m. K for pores of 5mm diameter. This last value is similar to the thermal conductivity of many refractory oxides and silicates, and an order of magnitude higher than for gas conduction.

For particulate structures, similar considerations provide an approximation for Kpefr-For example, for spheres KReff=k2.T3.d/Vf, (9) where k2 combines the Stefan-Boltzmann constant, emissivity and takes account of the radiation scattering cross section from particle geometry and dimensions, with a value of the order of 1 0~6W/m2. K4, d is particle diameter, and Vf is the particulate volume fraction. As a first approximation for refractory oxides, KRff is a minimum for spherical particles when their diameter, d, is ##/2, that is-1 urn at 1000°C and 0. 8, at 1500°C. Thus, a 50% volume fraction particulate structure consisting of 1 llm diameter refractory oxide spheres would have a predicted KRer, at 1 500°C of ~O. O l W/m. K.

Equations (4), (8) and (9) highlight the importance of micro-structural dimensions as determinants of thermal conductivity and, together with the basic theory, they inform how to affect low thermal conductivity at high temperatures : + Utilise the low thermal conductivity of gases by making highly porous materials.

Reduce solid conduction by minimising the volume fraction of the solid phase.

Employ materials of complex structure that have low intrinsic solid conductivity.

+ Minimise radiative and gas conductive heat transfer by introducing nano-and micro- metric dimensioned pores, particulates and other structural features.

In practice, the requirements for low thermal conductivity and low thermal mass have to be weighed against the requirements of the materials to be strong enough for a given application.

Accordingly compromises are reached.

In Figure 3 are plotted the bulk densities and 1000°C thermal conductivities of a wide range of alumina/silica/silicate refractories from various manufacturers. Also included is a line representing a"rule of mixtures"relationship K = 1. 5 [(1-Vd)/(l+Vd/2)], (10) where Vel is the volume fraction of void (air porosity), related to bulk density assuming the solid density is 3g/ml, and assuming that the solid Kc value at 1000°C is 1. 5W/m. K. These data illustrate many of the practical features of high temperature insulating materials.

Even the so-called"dense"refractories typically have porosity in the range 10 to 25% because they are processed from an assemblage of grains for which it is practically impossible to engineer ideal packing configurations that will sinter or bond fully dense. At these levels, porosity as a"second phase"has an effect on corrosion/erosion resistance and mechanical properties, but less of an effect on thermal conductivity than structural features of the solid (s). Thus, dense refractory concretes, that are complex structures of mixed solids, have lower thermal conductivities than fired (sintered) refractories, at similar bulk densities.

When porosity exceeds about 40% refractories are regarded as having useful insulating characteristics, and the thermal conductivity of the best of these porous structures obey the relationship of equation (10). Again, the insulating concretes typically have lower thermal conductivities than fired insulating firebrick (IFB), for the same reasons as the dense materials.

Porosity is the major determinant of"thermal mass". Thermal mass is the mass within a given volume which has to be heated to the application process temperature, and, perhaps, better expressed as volume heat capacity (J/m3. K), the product of specific heat of the dense solid and relative bulk density. In fact the effect of composition with the most widely used materials is small since the specific heat of alumina/silica/silicate solids is very similar.

Thermal mass is an important design consideration for intermittent and cyclic thermal processes, because the energy taken to raise the temperature of the refractory on heating cannot be recovered efficiently on cooling.

For groups of similar porous structure insulating refractories, it is the bulk density, determined by porosity, that dictates strength and the maximum service temperature, see Figure 4. Strength in these high porosity, flawed, materials is largely related to the volume fraction of interconnecting solid. The maximum service temperature is determined by the properties of the solid material, the degree of porosity, and the size of pores and particles making up the microstructure, all of which influence thermal stability. The significance of structural dimensions is described further below.

At the same porosity, the lowest thermal conductivity porous structures are those with the smallest median pore sizes, as shown in Figure 5. The contribution of conduction through the solid bridges between pores accounts for the thermal conductivity of real materials being considerably higher than that predicted for radiative heat transfer alone, equation (8).

Nevertheless, large pore size materials, usually foamed or bubble (hollow sphere) ceramics, are used in applications where low thermal mass is more important than conductivity, and, particularly, if higher strengths are needed. The greater strength of these materials is provided by the higher integrity interconnecting solids.

There are practical lower limits to pore sizes in these ceramic materials associated with their propensity to sinter when fired or used at high temperatures. The smaller the pore size, or for that matter a discrete solid phase, the greater are the driving forces for sintering to a lower specific surface energy state. The radiative conductivity benefits remain as long as the pores are maintained. Practically, it is probable that pore sizes and micro-structural features much less than 0. 1 um would be impossible to create and sustain at high temperatures.

The high temperature insulating materials with the lowest thermal conductivity combined with the lowest thermal mass are particulate structures. Fibrous insulating materials have the lowest thermal mass, and it is their unique combination of low thermal mass and low conductivity that has favoured them over insulating firebrick (Figure 6) for the majority of thermally cyclic applications.

At even lower thermal conductivity are the micro-porous materials. These are combinations of powders and fine fibres, which capitalise on every physical characteristic that can be employed to reduce heat transfer. And finally, having similar thermal conductivity as the fibrous materials but typically higher bulk densities are the natural and synthetic flake-like materials. Insulation materials manufactured from mica or exfoliated vermiculite flakes are well known.

For example, JP57022158 discloses a porous mica laminate comprising mica flakes, pulp fibre and PVA fibre, intended for use as electrical or heat insulation material. However, such a composition is unsuitable for use at refractory temperatures, due to combustion of the organic constituents. JP11157942 discloses a moulded vermiculite part comprising swollen vermiculite, peroxide and a resin binder, which yields a porous ceramic material upon firing, capable of withstanding high temperatures. However, the production method of such a material is time consuming and costly, involving two separate heating stages. GB485506 discloses an insulation material comprising exfoliated vermiculite, organic fibres, an inorganic binder, and in most cases, asbestos, causing serious health problems.

The common feature of these insulation materials is the use of minerals that occur naturally in a flake-like state. However, such naturally flaky materials are aggregated, with many flake layers in each particle, leading to the insulation properties being affected to some extent by the properties of the aggregated particle, and not just those of the flakes themselves.

Vermiculite may exfoliated to a certain degree to open up this aggregated structure and increase porosity, although this is not as effective as using individual flakes. None of these documents disclose the production of individual flakes for incorporation in an insulation material. A further group of insulating materials, for which data are not shown, are those based on carbon-powders, fibres and plates (exfoliated graphite). Having both useful dimensions and being close to black body emitters they have particularly low radiative conductivity. Their use in oxidising environments is restricted to temperatures below -500°C, but they are used in very high temperature controlled atmosphere furnaces and other special applications.

The thermal conductivity of these particulate structures is close to that predicted by modelling radiative heat transfer, Figure 7, because the particle dimensions and arrangements minimise the contribution of conduction through solid and gas phases. For the fibrous materials shown in Figure 3 thermal conductivity is lower at higher bulk densities, in agreement with theory, because of the increased volume fraction of the radiation scattering fibres.

The theoretical predictions shown in Figure 7 are for perfect particulate structures of uniform dimensions. In practical materials, such structures are not achieved, but this has its advantages as well as disadvantages. The data for fibre blanket insulation illustrate this.

First, high temperature vitreous fibres are melt processed and include, typically up to 50wt%, of non-fibrous particles, called"shot". These shot particles have dimensions in the range 10- 500go, and, therefore, contribute little to radiative scattering but, rather, increase conductivity by solid-solid heat transfer. Second, fibres that are melt-blown or spun are amorphous and have log-normal diameter distributions. The disordered amorphous structure has low solid conductivity, although this is lost when the material devitrifies at temperatures >-1000°C.

Typically, a fibre product with a median diameter of about 2um will contain fibre of diameter ranging from <1 um to >10um. This is a useful range for radiative scattering from ambient to high temperatures, because the radiation is spectral and the peak shifts to lower wavelengths as temperature increases according to Wien's law (equation (7)). (For this reason, semi-empirical relationships between conductivity and particulate dimensions suggest lower thermal conductivities than theory, e. g. W C Miller and T A Scripps, Ceramic Bulletin, 1982, vol. 61 (7), pp. 711-724.) Generally, these powder and fibrous materials are"particles dispersed in a gas"and have no useful structural integrity unless a third phase"binder"is introduced. Even for these bound products, structural integrity is very low. Fibrous structures can be afforded some cohesion by interweaving the fibres which are frequently several cms long, and for this reason these products are commonly referred to as"wools". At the other extreme, some micro-porous products may only be used when contained within a bag woven from high temperature fibres.

A further factor that is increasingly becoming of relevance to choice of materials is hazard assessment.

Since 1997 Member States of the EU have introduced a hazard classification scheme for man-made vitreous silicate fibres (MMVFs), which includes the majority of fibrous insulation materials. This scheme classifies some MMVFs as probable human carcinogens by inhalation.

Also in 1997 the World Health Organisation's International Agency for Research on Cancer (IARC) published its evaluation that certain occupational dusts containing crystalline silica, which is a common component or reaction phase in insulating refractories, are human carcinogens.

The consequences of these decisions in Europe are that many thermal insulation products containing MMVFs will be supplied with warning labels. There is no requirement in Europe to label crystalline silica-containing products, although there is in other jurisdictions. Users may be required to undertake risk assessments when using any of these products. Producers, suppliers and user trade associations will be able to provide further, detailed guidance.

The hazard assessments are primarily concerned with respiratory disease and are based on precautionary approaches to the results of tests on animals, with limited or no human (epidemiological) data. It has to be stressed that for MMVFs, there is no association between MMVF exposure and respiratory disease in human populations exposed for over 40 years. For crystalline silica, the IARC report notes that"carcinogenicity in humans was not detected in all industrial circumstances studied"and"may be dependent on inherent characteristics of the crystalline silica or on external factors affecting its biological activity".

Most human respiratory diseases caused by dust demonstrate clear dose-response, or exposure-response, relationships and some have no observable adverse effect levels (NOAEL) of dose/exposure, below which there is no (detectable) disease incidence. Dose is usually defined as the burden of dust in the deep lung, and is determined by the rates of deposition and clearance of the specific dust in question. Deposition is largely determined by the exposure concentration and the dust particle size and shape. To reach the deep lung (to be respirable) particles need to be very small, and deposition efficiencies are typically low, for example for fibres <10%.

Clearance of dust from the lung is determined by its chemical characteristics and by size.

Most inert dusts are readily removed from the deep lung by the body's natural defence mechanisms, but these may be defeated if the dust is chemically toxic or if they are overloaded by exposure to a very high respirable dust concentration for a sufficient period of time.

Many positive outcomes to dust exposures in animal experiments are now thought to have been caused by «'overload", an (certain human respiratory conditions are considered to result only from very high concentration exposures, even if for a short period of time. In the absence of overload, the characteristic clearance rate of the dust has been called its "biopersistence". Thus, the lower the biopersistence the faster the clearance rate and the lower the steady-state dose.

These factors provide general guidance on reducing/eliminating the risk to human health from exposure to dusts generated from high temperature insulating refractories : Reduce exposure concentration and time to the lowest practicable. Regulatory exposure limits or guidelines operate in most jurisdictions, and should be met where necessary through use of engineering controls or personal protective equipment.

These limits are mostly stated as time-weighted averages. However, it is prudent to avoid even short-term exposures above the limits.

+ Evaluate the use low biopersistence alternative products where such a choice is available. MMVF based thermal insulation manufacturers have developed and brought to market such products over the past 10 years. For some applications above 1000°C low biopersistence calcium-magnesium-silicate wools are available as alternatives to the established refractory ceramic fibre products.

Quantitative risk assessment is a developing science that seeks to relate excess lifetime risk to specific exposure scenarios. For MMVFs, the most recent studies show that for long term occupational exposures, at concentrations about double current industry standards, excess lifetime risk is between 2x10-5 and lxlO4, about the same risk as dying from the results of a bee sting.

Similar quantitative risk assessments are not yet available for exposures to crystalline silica- containing dusts, because the dosimetry modelling is more complex. Nevertheless, many who have studied the problem have concluded that control measures for the prevention of silicosis will be effective in reducing any excess in lung cancer, and that the risks of silicosis are contained by avoiding exposures to crystalline silica greater than 0. lmg/m3.

The inventors have realised that there is a class of materials that provide :- low thermal conductivity-equivalent to or better than that of fibrous insulation low thermal mass higher strength than fibrous insulation without the need to use fibrous materials with their perceived hazards.

Accordingly the present invention provides a porous composite body comprising a plurality of flakes bonded by an inorganic binder, the flakes having a thickness of less than 10tm and the porous composite body having a porosity of > 40%.

An advantage of the present invention is that specific insulation properties may be engineered by the production of individual flakes of specified composition, dimensions or curl.

The invention will be further described in the following with reference to the drawings in which :- Figure 1 shows typical energy losses for intermittent furnace operation with different insulating refractories Figure 2 shows insulating mass per unit area for typical furnace walls with different insulating refractories Figure 3 plots thermal conductivity, bulk density and porosity of alumina/silica/silicate refractories Figure 4 plots thermal conductivity, strength and maximum service temperature against density for insulating firebrick Figure 5 plots thermal conductivity against pore size for porous insulating refractories Figure 6 plots thermal conductivity and thermal mass of fibre and firebrick insulating refractories Figure 7 plots thermal conductivity against structural dimensions for particulate insulating refractories Figure 8 plots model data for the thermal conductivity of alumina flake particulate insulating refractories varying the assumed absorption values of the alumina Figure 9 plots model data for the thermal conductivity of alumina flake particulate insulating refractories varying the density of the refractory Figure 10 plots model data for the thermal conductivity of alumina flake particulate insulating refractories varying the thickness of the flakes Figure 11 plots thermal conductivity against temperature for Superwool 607TM fibre and flake refractory insulation materials of corresponding composition Figure 12 plots thermal conductivity against temperature for alumina flake refractory insulation material.

Computer modelling was undertaken by the applicants based upon the above mentioned criteria. The model looked to the thermal conductivity of mullite and alumina flakes. The model attempted to take into account the effect of heat barrier resistances associated with porous grain boundaries and microcracks. Different values for scattering/absorption factors were used based on data measured on a thermal conductivity rig. The equation used for the effective thermal conductivity is as follows :- where :- ks = Solid phase conduction Xrad= Thermal conductivity from radiation Pore phase conduction (air) and II = porosity The value assigned to the absorption coefficient in the solid phase conduction calculations has a great effect on the whole model and values are displayed in Figure 8 where the flake thickness and board density remain constant at 10m and 1000kg/m3 respectively. The absorption coefficient was varied from 0. 4 up to 0. 85 and resulting values compared to actual values measured on alumina flake board (F13 in Figs 8 to 10 which is a board produced from alumina flakes of <500tm thickness).

Figure 9 represents thermal conductivity values for alumina flake boards of different density (600, 800, 1000 and 1200 kg/m3 respectively) with different flake thicknesses, but maintaining the absorption coefficient at 0. 85. The density has an important effect on thermal conductivity. At low temperatures the conductance plays an important part and low density material gives lower values, but as the temperature increases then the radiative effect becomes more dominant and the figure indicates that higher densities will result in lower thermal conductivities.

In Figure 10 the density has been kept constant at l000kg/m3 but the flake thickness has been varied (l, um, 5, um, lO, um and 20pm respectively). This shows that the thinner flakes should result in lower thermal conductivities at all temperatures and emphasises that the thickness of flakes needs to be as small as possible with calculated values for the I Ogm to the 20gm thick flakes doubling the thermal conductivity at 1400°C. Comparison with the actual measured values for the thick flakes ofF13 is instructive. At thicknesses of about 500um the thermal conductivity varies almost linearly with temperature, at the small flake thicknesses of lO, Lm or less a more complex relationship applies and lower thermal conductivities are shown in the range 600°C upwards, the very region in which refractories are required to work.

This is of course an oversimplification.

For example, the FI 3 measured values appear better than the theoretical figure. This is possibly because the real alumina flakes are not the idealised solids used in the theory and, as mentioned above, complexity reduces thermal conductivity.

Also, if the flakes are too thin problems with transparency of the flakes to radiation arise which in the real world would lead to increase in thermal conductivity for very thin flakes.

However there are benefits to use of flakes other than those predicted by these calculations.

Referring to Fig. 7, it was mentioned above that the actual thermal conductivity of fibre blanket is higher than that predicted from theory, largely because of the presence of shot, which is a product of the fibre manufacturing process. Processes for forming flake (as is explained below) tend to produce less or no shot and so the performance of flake insulation is likely to approach that of theory more than fibre insulation does. Furthermore, a small degree of curl in the flakes produces an increase in porosity, again producing insulating performance closer to that predicted by theory.

Additionally, because flakes a-e not fibres they are less likely to attract health concerns than do fibres at present.

The following examples show these principles in practice.

Example 1 The applicants have made board from glass flakes to test this theory. The glass flakes used were obtained from Glassflake Limited of Leeds, England who hold European Patent No. 0289240 to a method and apparatus for producing glass flake. This method allows the manufacture of glass flake of precisely controlled thickness. The glass flake is conventionally used as additives for paints, additives for thermoplastics, and for producing filter media.

Flake is made to various grades and examples of the particle size ranges available are given below :- Grade (s) Thickness Particle size distribution GF200 1. 5 µm >1700 µm 0 GF300 2. 5 Fm 1700 to 150 µm 80% or more GF400 4. 5 tim 150to<50pm20% or less GF200M 1. 5 µm >1000 µm 0 GF300M 2. 5 Am 1000 to 300 µm 10% or less GF400M 4. 5 pm 300 to 50 µm 65% or more <50 Fm 25% or less GF005 3.5-5. 5 µm >150 µm 2% or less GF002 1. 9-2. 5 um 0 50 to 150 llm 10% or less < µm 88% or more These grades have a typical glass composition (wt%) :- Si02 64-70% Ti02 0-1% A1203 3-6% CaO 3-7% MgO 1-4% Na20 8-13% ZnO 1-5% B203 2-5% K20 0-3% The flake is manufactured by discharging a melt stream (6-12mm diameter) at 1250°C into a highly polished stainless steel spinning cup, with the side walls at an angle of-45°. The glass quenches, firstly depositing a thin solid layer and subsequent material is drawn up the side walls by centrifugal force and is then displaced as a film over the edge. Material is drawn out between two parallel venturi plates within a cyclone chamber, therefore varying the air velocity, resulting in flakes from 1-Sum controlled thickness.

Boards were made using different techniques to assess flake insulation. Samples of 1. 5, 2. 5 and 4. 5pm thick flakes were obtained (Grades GF200, 300 and 400) and vacuum cast into preformed boards,. Approximately 60g of glass flake material were dispersed in l litre of cationic starch solution (0. 5%) containing several grams of a common commercial colloidal silica e. g. Nyacol 1430. Once mixed in sufficiently by hand the material was transferred to a deep casting tool of the dimensions 160 x 11 Omm and a vacuum applied, drawing the excess liquid through a nylon mesh and retaining the flake material. The preform was demoulded and dried in an oven typically at 50-80°C, resulting in a board of net shape and ~ 10- 20mm thickness.

The board had a typical density of 0. 13g/cm3. Thermal conductivity tests on this low softening-point (650°C) C glass board were carried out up to 600°C and gave values of-0. 17 W/mK which is similar to that of commercial 128 grade KaowoolT blanket (density 0. 128g/cm). The boards however have the advantage that they are self-supporting.

The applicants did investigate the use of the finer grades of Glassflake (GF200M, 300M and 400M) but found that they posed problems in vacuum casting in that removal of water was difficult. A preferred range of sizes is that at least 50% of the flakes have a length of greater than 50pu, more preferably greater than 150um. Still more preferably, at least 80% of the flakes have a length of greater than 150, um.

Example 2 Alternative processing routes permit alternative products to be made. A viscous processing route was investigated. A board was made using the following recipe and by the method indicated below.

Recipe :- Tap Water 74. 61% Natrosol 250 HHRTM 1. 57% Nyacol 1430TM 1. 57% GF400TM 22. 15% FairyTM 0. 10% The ingredients other than water were :- Ingredient Nature and source Natrosol 250HHR Water-soluble polymer-Hercules hydroxyethyl cellulose (HEC) Nyacol 1430TM Sodium-stabilised colloidal silica DuPont dispersion-30% solids GF400TM Glass flake as specified above Glassflake Limited Fairy Domestic washing-up liquid The polymer (Natrosol 250HHRTM) served to produce the necessary highly viscous solution when dissolved in the water, and also acts as a low temperature binder. This polymer has a delayed solubility feature, enabling it to be dissolved smoothly and completely. Other such polymers dissolve almost immediately but this leads to the production of so-called"fish eyes" -agglomerates of undissolved polymer in a gelatinous coat. Such fish eyes require special techniques or mixers to either avoid their production or disperse them.

The colloidal silica (Nyacol 1430TM) functions in two ways-firstly as a high temperature binder ; secondly it serves to further thicken the aqueous polymer solution.

The domestic washing-up liquid served as a foaming agent but can be dispensed with to achieve higher densities.

The basic stages of mix preparation were to dissolve the polymer in water, add colloidal silica, and finally to disperse the glass flake. The mixture produced, which had a pasty consistency, was rolled to produce a board.

The board had a density of 0. 24g/cm3. Higher densities are achievable by this route, for example by omitting the foaming agent.

Compositions of this type can be extruded through a die or into a mould to form ceramic bodies. In this specification the term extrusion is to be taken as including methods such as rolling, where the die is in motion, or injection moulding, where material is forced into a mould.

This route also enables hydrophobic or low density materials to be incorporated into a board.

Exfoliated vermiculite is a hydrothermally treated mica comprising thin flakes. Boards of this material have been made by pressing vermiculite with a binder to form low density boards for thermal and acoustic insulation.

Although the glass flake boards were usable up to 600°C, use at higher temperatures is not possible due to the melting point of the glass from which they were made. Higher temperature materials are used for the manufacture of fibre insulation. Any material that can be made into a fibre can in principle be made into flake by the process of European Patent No. 0289240. Typical fibres that are used for high temperature applications include alumino- silicate fibres and the alkaline earth silicate fibres (see for example W087/05007, W089/12032, W093/15028, W094/15883, W096/02478, and W097/49643).

Moreover, the requirements for producing a fibre (in terms of viscosity) are much more demanding than those for producing flake and so a wider range of chemistries may be used for flake production than for fibre production.

The conventional processes for producing fibre disperse a stream of molten composition either by either blowing the stream or allowing it to impinge on a rapidly rotating wheel.

Droplets of composition are flung out and trail a tail of composition which becomes the fibre.

The droplets form what is known as shot. The analogous process for producing flake from the melt does not require such droplets and so the glass flake is relatively shot free.

Example 3 The applicants have also investigated the use of flaked ceramic materials for forming boards with improved insulation properties over different temperature ranges. Batches of flake for insulation use were produced according to the process outlined in Example 1 from a material of similar composition to Superwool 607 MAXTM fibre insulation. (Superwool 607 MAXTM has a nominal composition (by weight) of Si02 65%, CaO 20%, MgO 15% with a tolerance of about 0. 5wt% on each component). The resultant flakes showed some degree of curl. Typical flake thickness was less than 1. 5µm. Such boards were intended to be a high-temperature analogue of those described in Example 1.

The composition of the batch, as given below, was determined using X-ray fluorescence. % Oxide Flake material Na20 0. 05 MgO 15. 45 Al2O3 0.05 Si02 64. 2 P2O5 <0. 05 SO3 <0. 05 K20 <0. 05 CaO 19.85 Ti02 <0. 05 V205 <0. 05 Cr203 <0. 05 Mn304 <0. 05 Fe2O3 0.06 ZnO <0. 05 SrO <0. 05 Y2O3 ZrO2 <0. 05 BaO <0. 05 HfD2 <0. 05 LOI@^1250°C N/A Total 99. 56 In order to test the thermal conductivity and shrinkage characteristics, test preforms were manufactured from the batch. 30g of flake material were dispersed in 1 litre of anionic starch solution (e. g. 1. 5g/l of Wisprafloc) containing 1. 5 g of anionic commercial colloidal silica (e. g. Nyacol 1430). Once mixed sufficiently by hand, the suspension was poured into a vacuum casting tool measuring 65mm x 120mm attached to a vacuum pump, using slight hand pressure whilst the liquid was sucked into the collection flask. Once de-moulded the preform was allowed to dry on a stainless steel mesh support, with a piece of nylon mesh to avoid adhesion once dry, and then placed in a drying oven (between 80 and 120°C) overnight. Typical density of these preforms was 0. 16g/cm3.

Figure 11 shows the results of thermal conductivity tests over a range of temperatures for the Superwool 607 MAXT preforms (pressed in formation to get a higher density), in comparison with Superwool 607 MAX TM fibre. The flaked Superwool material shows a much lower thermal conductivity than the fibre at higher temperatures, approximately 0. 20W. m- IK-l compared with 0. 27W. m-'. K~1 at 1000°C and 0. 27W. m-'. Ksi compared with 0. 35W. m-'. K-' at 1200°C.

As a direct comparison can be made to fibre boards of the same composition, tests were also carried out to determine the shrinkage of preforms made from the flaked Superwool material.

The dimensions of the preforms after 24 hours at each temperature were compared with those of the as-made preforms in order to determine the amount of shrinkage. Results from both consecutive testing (same sample retested at the next test temperature) and sequential testing (new sample tested for each test temperature) are shown below : Shrinkage temperature Linear shrinkage per test type (%) °C Consecutive Sequential 1200 4. 1 4. 5 1260 4. 3 3. 0 1300 4. 6 5. 6 The variation between the values is most likely to be due to a sintering effect on the surface of the material during testing. A skin is formed on the surface of the material during consecutive testing, which may inhibit shrinkage. However, the flakes in the interior of the sample retain their origin form and are unaffected. The shrinkage is broadly similar to boards manufactured from fibres of the same composition.

Example 4 In order to demonstrate the extent to which the use of the flaked material may be tailored to meet particular insulation requirements, the applicants fabricated an alumina preform, with a density of 900kg/m3 and porosity of 79%. The alumina flakes were produced from a precursor sol using aluminium nitrate hydrate (Al (NO3) 3. 9H20-AN in table below) and aluminium isopropoxide (AIP in table below) using fluoride ion doping. This promotes preferential grain growth along one crystalline axis, producing a thin plate-like structure. The composition of the sol was as follows : Mass of Reagent Composition (wt%) AN AIP NH4F Water Al203 | F- 900. 00 979. 20 23. 00 2988. 00 96. 99 3.

Following the mixing stage, any alcoholic product and water were removed via rotary evaporation at 40°C, to produce a viscous sol with a treacle-like appearance. The material was then fired at 1200°C for 2 hours to convert the material to oxide, resulting in the growth of flakes with an average thickness of lum. The flakes were then incorporated into a preform via vacuum forming (as in Example 2) according to the following recipe : Cerasol (alumina sol 20% w/w) 280g CaCl2. 2H20 15g Solvatose (anionic starch) 0. 5% 500g Alumina flake 200g Water 2180g The alumina sol and calcium chloride were mixed first in a shear mixer, followed by addition of the alumina flake.

The commercial alumina boehmite (AIO (OH)) sol used [Cerasol (Alcan Chemicals, UK)] is conventionally used as a binder in ceramic and refractory products. The applicants have found that such a binder containing for example consisting of 20% w/w equivalent of alumina may be treated with other salts of either calcium or strontium and possibly other group I and II compounds to form a precursor aluminate sol.

Typically >I % equivalent to the oxide of calcium or strontium is added by means of a shear mixer allowing homogeneous dispersion, resulting in an increase in viscosity of the treated sol. In this case sufficient group II salt in the form of the chlorides were added to give the stoichiometry-5A1203. MO.

The material can then be fired to convert the material to an oxide, in this case a firing temperature of 1250°C was used to produce predominantly aluminate phases.

SEM results show that the reaction between the fine boehmite colloids and the group II salt produces flakes with a thickness typically <0. 5pm and several microns long. A composition of stoichiometry 6AI203. CaO gave thick flakes whereas one of, for example, 4. 4A1203. Ca0 gave fine flakes (e. g. 25im thickness). Similar controlled production of flake can be achieved using fluoride ions to agglomerate alumina sols.

This preferential growth along a crystal axis is however not seen when a coarser grade of boehmite sol is used, eg. Bacosol (Alcan Chemicals, UK). The reason is that reaction at low temperature requires a large surface area and hence fine particles. The increased homogeneity that results from using a fine sol and a salt as starting materials, and the low reaction temperature resulting, permits the directional grain growth necessary to form a thin flake.

Cerasol is a fine, typically 50nm colloid of AIO (OH) (boehmite), and when mixed with salt (CaCl2 in this case) an increase in viscosity is observed implying electrostatic attraction leading to partial flocculation. Bacasol has a larger particle distribution, some fine (50nm) and up to 500nm (0. 5 microns), and an increase in viscosity is not observed during mixing.

The fine flake materials may u ell be useful as low density binders which obtain a distinctly flake like structure (both for flake and fibrous materials), consisting of flakes with a thickness typically <0. zum several riicrons long. These materials would be particularly suited in applications where they bind materials of like composition eg. Calcium or strontium aluminate fibres or flakes.

Such a production method also allows dimensional control of the flakes via further heat treatments.

Figure 12 shows the thermal conductivity over the temperature range 100 to 1200°C of aboard made to the above recipe. Initially the thermal conductivity is higher than that for both flake and fibre Superwool materials, but shows a sharp drop in the region of 800°C, with a minimum thermal conductivity of 0. 27 at 1000°C. This makes the material extremely suitable for use at refractory temperatures. The variation in thermal conductivity is most likely due to the density of the board and the behaviour of the alumina flake material creating an unusual combination of thermal transport mechanisms over the desired temperature range.

By making flake materials of different composition the full range of refractory requirements from the low hundreds up to and beyond 1600°C are achievable.

The flake used in the above experiments did not consist of flat sheets. A degree of curl is inevitable in the production process. This may be advantageous in processing terms in allowing channels for the removal of liquid from the ceramic in processing. The composition, dimensions and degree of curl of individual flakes may be specified and engineered, allowing the full range of refractory requirements from the low hundreds up to and beyond 1600°C to be achieved.

Although the above examples only illustrate vacuum forming and extrusion, the application is not limited to extrusion or vacuum forming, but includes other techniques that provide inorganic bonds, such as freeze-casting.